How do monoamines shape what we feel, decide, and experience — and can we measure them in real time?
I develop machine learning models for real-time neurochemical inference from electrochemical data acquired via intracranial electrodes in human patients. My focus is on building deep learning architectures that generalize neurotransmitter concentration estimates across probes, labs, and recording conditions. The direction is multimodal — fusing neurochemical data with electrophysiology and behavior. Ultimately, this opens the door to investigating how neurotransmitter dynamics shape human phenomenology — emotion, subjective experience, and conscious state transitions that are difficult or impossible to model outside human subjects.
The neurochemistry is my latest approach to a question I've pursued for over a decade — Bayesian neural source localization, neural markers of awareness in infants, the influence of expectations on unconscious processing, and the mathematical formalization of Integrated Information Theory (IIT), a theory of consciousness.
Deep learning models (InceptionTime, meta-encoders) that decode sub-second monoamine dynamics — dopamine, serotonin, norepinephrine — from electrochemical recordings in human patients. The goal is probe-invariant, lab-invariant inference that works across electrode types and recording conditions, moving toward multimodal integration of neurochemical, electrophysiological, and behavioral data.
Mathematical foundations of Integrated Information Theory. Uniqueness proofs for intrinsic information measures using functional equations and Minkowski-type inequalities. Earlier work on Bayesian source localization of neural signals, unconscious cognitive processing, predictive coding, and neural markers of perceptual awareness during sleep and in infants.
All publications on Google Scholar · 1,200+ citations · 20 publications